An Anthropomorphic Motion Retargeting Method Based on an Adaptive Genetic Algorithm

Abstract

A manipulator with human-like movements enables better human-machine interaction and a more intuitive experience for the operator when he or she teleoperates it. In this work, we developed an interactive interface based on the robot operating system (ROS). An adaptive genetic algorithm was used to find the optimal robot arm configuration based on the positions of the elbow and wrist of the human arm inputted by the Kinect camera and inertial measurement unit (IMU). Validating the feasibility of the proposed method were the fixed pose experiment, the continuous wave experiment, and the pick-place experiment. Comparison of the proposed method to the conventional inverse kinematics algorithm and the general genetic algorithm shows that the wrist position error obtained by the three methods is smaller, and the elbow position obtained by the proposed method is closer to the target position. Based on the proposed method, the polygonal area enclosed by the configurations of the operator's arm and the manipulator is smaller, and the anthropomorphism is better.